CoreSight September 2020

coresight@lists.linaro.org

21 participants
34 discussions

[PATCH v2] coresight: etm4x: Handle unreachable sink in perf mode

by Suzuki K Poulose

If the specified/hinted sink is not reachable from a subset of the CPUs, we could end up unable to trace the event on those CPUs. This is the best effort we could do until we support 1:1 configurations. Fail gracefully in such cases avoiding a WARN_ON, which can be easily triggered by the user on certain platforms (Arm N1SDP), with the following trace paths : CPU0 \ -- Funnel0 --> ETF0 --> / \ CPU1 \ MainFunnel CPU2 / \ / -- Funnel1 --> ETF1 --> / CPU1 $ perf record --per-thread -e cs_etm/@ETF0/u -- <app> could trigger the following WARNING, when the event is scheduled on CPU2. [10919.513250] ------------[ cut here ]------------ [10919.517861] WARNING: CPU: 2 PID: 24021 at drivers/hwtracing/coresight/coresight-etm-perf.c:316 etm_event_start+0xf8/0x100 ... [10919.564403] CPU: 2 PID: 24021 Comm: perf Not tainted 5.8.0+ #24 [10919.570308] pstate: 80400089 (Nzcv daIf +PAN -UAO BTYPE=--) [10919.575865] pc : etm_event_start+0xf8/0x100 [10919.580034] lr : etm_event_start+0x80/0x100 [10919.584202] sp : fffffe001932f940 [10919.587502] x29: fffffe001932f940 x28: fffffc834995f800 [10919.592799] x27: 0000000000000000 x26: fffffe0011f3ced0 [10919.598095] x25: fffffc837fce244c x24: fffffc837fce2448 [10919.603391] x23: 0000000000000002 x22: fffffc8353529c00 [10919.608688] x21: fffffc835bb31000 x20: 0000000000000000 [10919.613984] x19: fffffc837fcdcc70 x18: 0000000000000000 [10919.619281] x17: 0000000000000000 x16: 0000000000000000 [10919.624577] x15: 0000000000000000 x14: 00000000000009f8 [10919.629874] x13: 00000000000009f8 x12: 0000000000000018 [10919.635170] x11: 0000000000000000 x10: 0000000000000000 [10919.640467] x9 : fffffe00108cd168 x8 : 0000000000000000 [10919.645763] x7 : 0000000000000020 x6 : 0000000000000001 [10919.651059] x5 : 0000000000000002 x4 : 0000000000000001 [10919.656356] x3 : 0000000000000000 x2 : 0000000000000000 [10919.661652] x1 : fffffe836eb40000 x0 : 0000000000000000 [10919.666949] Call trace: [10919.669382] etm_event_start+0xf8/0x100 [10919.673203] etm_event_add+0x40/0x60 [10919.676765] event_sched_in.isra.134+0xcc/0x210 [10919.681281] merge_sched_in+0xb0/0x2a8 [10919.685017] visit_groups_merge.constprop.140+0x15c/0x4b8 [10919.690400] ctx_sched_in+0x15c/0x170 [10919.694048] perf_event_sched_in+0x6c/0xa0 [10919.698130] ctx_resched+0x60/0xa0 [10919.701517] perf_event_exec+0x288/0x2f0 [10919.705425] begin_new_exec+0x4c8/0xf58 [10919.709247] load_elf_binary+0x66c/0xf30 [10919.713155] exec_binprm+0x15c/0x450 [10919.716716] __do_execve_file+0x508/0x748 [10919.720711] __arm64_sys_execve+0x40/0x50 [10919.724707] do_el0_svc+0xf4/0x1b8 [10919.728095] el0_sync_handler+0xf8/0x124 [10919.732003] el0_sync+0x140/0x180 Even though we don't support using separate sinks for the ETMs yet (e.g, for 1:1 configurations), we should at least honor the user's choice and handle the limitations gracefully, by simply skipping the tracing on ETMs which can't reach the requested sink. Fixes: f9d81a657bb8 ("coresight: perf: Allow tracing on hotplugged CPUs") Reported-by: Jeremy Linton <jeremy.linton(a)arm.com> Tested-by: Jeremy Linton <jeremy.linton(a)arm.com> Cc: Mathieu Poirier <mathieu.poirier(a)linaro.org> Cc: Mike Leach <mike.leach(a)linaro.org> Signed-off-by: Suzuki K Poulose <suzuki.poulose(a)arm.com> --- Change since v1: - Updated the description, added Tested-by. - No code changes - Rebased on coresight/next --- drivers/hwtracing/coresight/coresight-etm-perf.c | 10 ++++++++++ 1 file changed, 10 insertions(+) diff --git a/drivers/hwtracing/coresight/coresight-etm-perf.c b/drivers/hwtracing/coresight/coresight-etm-perf.c index 1a3169e69bb1..9d61a71da96f 100644 --- a/drivers/hwtracing/coresight/coresight-etm-perf.c +++ b/drivers/hwtracing/coresight/coresight-etm-perf.c @@ -321,6 +321,16 @@ static void etm_event_start(struct perf_event *event, int flags) if (!event_data) goto fail; + /* + * Check if this ETM is allowed to trace, as decided + * at etm_setup_aux(). This could be due to an unreachable + * sink from this ETM. We can't do much in this case if + * the sink was specified or hinted to the driver. For + * now, simply don't record anything on this ETM. + */ + if (!cpumask_test_cpu(cpu, &event_data->mask)) + goto fail_end_stop; + path = etm_event_cpu_path(event_data, cpu); /* We need a sink, no need to continue without one */ sink = coresight_get_sink(path); -- 2.24.1

4 years, 3 months

[PATCH v2] coresight: cti: write regsiters directly in cti_enable_hw()

by Tingwei Zhang

Deadlock as below is triggered by one CPU holds drvdata->spinlock and calls cti_enable_hw(). Smp_call_function_single() is called in cti_enable_hw() and tries to let another CPU write CTI registers. That CPU is trying to get drvdata->spinlock in cti_cpu_pm_notify() and doesn't response to IPI from smp_call_function_single(). [ 988.335937] CPU: 6 PID: 10258 Comm: sh Tainted: G W L 5.8.0-rc6-mainline-16783-gc38daa79b26b-dirty #1 [ 988.346364] Hardware name: Thundercomm Dragonboard 845c (DT) [ 988.352073] pstate: 20400005 (nzCv daif +PAN -UAO BTYPE=--) [ 988.357689] pc : smp_call_function_single+0x158/0x1b8 [ 988.362782] lr : smp_call_function_single+0x124/0x1b8 ... [ 988.451638] Call trace: [ 988.454119] smp_call_function_single+0x158/0x1b8 [ 988.458866] cti_enable+0xb4/0xf8 [coresight_cti] [ 988.463618] coresight_control_assoc_ectdev+0x6c/0x128 [coresight] [ 988.469855] coresight_enable+0x1f0/0x364 [coresight] [ 988.474957] enable_source_store+0x5c/0x9c [coresight] [ 988.480140] dev_attr_store+0x14/0x28 [ 988.483839] sysfs_kf_write+0x38/0x4c [ 988.487532] kernfs_fop_write+0x1c0/0x2b0 [ 988.491585] vfs_write+0xfc/0x300 [ 988.494931] ksys_write+0x78/0xe0 [ 988.498283] __arm64_sys_write+0x18/0x20 [ 988.502240] el0_svc_common+0x98/0x160 [ 988.506024] do_el0_svc+0x78/0x80 [ 988.509377] el0_sync_handler+0xd4/0x270 [ 988.513337] el0_sync+0x164/0x180 This change write CTI registers directly in cti_enable_hw(). Config->hw_powered has been checked to be true with spinlock holded. CTI is powered and can be programmed until spinlock is released. Fixes: 6a0953ce7de9 ("coresight: cti: Add CPU idle pm notifer to CTI devices") Signed-off-by: Tingwei Zhang <tingwei(a)codeaurora.org> --- drivers/hwtracing/coresight/coresight-cti.c | 24 +++++---------------- 1 file changed, 5 insertions(+), 19 deletions(-) diff --git a/drivers/hwtracing/coresight/coresight-cti.c b/drivers/hwtracing/coresight/coresight-cti.c index c4e9cc7034ab..d04181c12a7f 100644 --- a/drivers/hwtracing/coresight/coresight-cti.c +++ b/drivers/hwtracing/coresight/coresight-cti.c @@ -86,22 +86,16 @@ void cti_write_all_hw_regs(struct cti_drvdata *drvdata) CS_LOCK(drvdata->base); } -static void cti_enable_hw_smp_call(void *info) -{ - struct cti_drvdata *drvdata = info; - - cti_write_all_hw_regs(drvdata); -} - /* write regs to hardware and enable */ static int cti_enable_hw(struct cti_drvdata *drvdata) { struct cti_config *config = &drvdata->config; struct device *dev = &drvdata->csdev->dev; int rc = 0; + unsigned long flags; pm_runtime_get_sync(dev->parent); - spin_lock(&drvdata->spinlock); + spin_lock_irqsave(&drvdata->spinlock, flags); /* no need to do anything if enabled or unpowered*/ if (config->hw_enabled || !config->hw_powered) @@ -112,19 +106,11 @@ static int cti_enable_hw(struct cti_drvdata *drvdata) if (rc) goto cti_err_not_enabled; - if (drvdata->ctidev.cpu >= 0) { - rc = smp_call_function_single(drvdata->ctidev.cpu, - cti_enable_hw_smp_call, - drvdata, 1); - if (rc) - goto cti_err_not_enabled; - } else { - cti_write_all_hw_regs(drvdata); - } + cti_write_all_hw_regs(drvdata); config->hw_enabled = true; atomic_inc(&drvdata->config.enable_req_count); - spin_unlock(&drvdata->spinlock); + spin_unlock_irqrestore(&drvdata->spinlock, flags); return rc; cti_state_unchanged: @@ -132,7 +118,7 @@ static int cti_enable_hw(struct cti_drvdata *drvdata) /* cannot enable due to error */ cti_err_not_enabled: - spin_unlock(&drvdata->spinlock); + spin_unlock_irqrestore(&drvdata->spinlock, flags); pm_runtime_put(dev->parent); return rc; } -- The Qualcomm Innovation Center, Inc. is a member of the Code Aurora Forum, a Linux Foundation Collaborative Project

4 years, 3 months

[PATCH] coresight: etm4x: Handle unreachable sink in perf mode

by Suzuki K Poulose

If the specified/hinted sink is not reachable from a subset of the CPUs, we could end up unable to trace the event on those CPUs. This is the best effort we could do until we support 1:1 configurations. Fail gracefully in such cases avoiding a WARN_ON, which can be easily triggered by the user on certain platforms, like : [10919.513250] ------------[ cut here ]------------ [10919.517861] WARNING: CPU: 2 PID: 24021 at drivers/hwtracing/coresight/coresight-etm-perf.c:316 etm_event_start+0xf8/0x100 ... [10919.564403] CPU: 2 PID: 24021 Comm: perf Not tainted 5.8.0+ #24 [10919.570308] pstate: 80400089 (Nzcv daIf +PAN -UAO BTYPE=--) [10919.575865] pc : etm_event_start+0xf8/0x100 [10919.580034] lr : etm_event_start+0x80/0x100 [10919.584202] sp : fffffe001932f940 [10919.587502] x29: fffffe001932f940 x28: fffffc834995f800 [10919.592799] x27: 0000000000000000 x26: fffffe0011f3ced0 [10919.598095] x25: fffffc837fce244c x24: fffffc837fce2448 [10919.603391] x23: 0000000000000002 x22: fffffc8353529c00 [10919.608688] x21: fffffc835bb31000 x20: 0000000000000000 [10919.613984] x19: fffffc837fcdcc70 x18: 0000000000000000 [10919.619281] x17: 0000000000000000 x16: 0000000000000000 [10919.624577] x15: 0000000000000000 x14: 00000000000009f8 [10919.629874] x13: 00000000000009f8 x12: 0000000000000018 [10919.635170] x11: 0000000000000000 x10: 0000000000000000 [10919.640467] x9 : fffffe00108cd168 x8 : 0000000000000000 [10919.645763] x7 : 0000000000000020 x6 : 0000000000000001 [10919.651059] x5 : 0000000000000002 x4 : 0000000000000001 [10919.656356] x3 : 0000000000000000 x2 : 0000000000000000 [10919.661652] x1 : fffffe836eb40000 x0 : 0000000000000000 [10919.666949] Call trace: [10919.669382] etm_event_start+0xf8/0x100 [10919.673203] etm_event_add+0x40/0x60 [10919.676765] event_sched_in.isra.134+0xcc/0x210 [10919.681281] merge_sched_in+0xb0/0x2a8 [10919.685017] visit_groups_merge.constprop.140+0x15c/0x4b8 [10919.690400] ctx_sched_in+0x15c/0x170 [10919.694048] perf_event_sched_in+0x6c/0xa0 [10919.698130] ctx_resched+0x60/0xa0 [10919.701517] perf_event_exec+0x288/0x2f0 [10919.705425] begin_new_exec+0x4c8/0xf58 [10919.709247] load_elf_binary+0x66c/0xf30 [10919.713155] exec_binprm+0x15c/0x450 [10919.716716] __do_execve_file+0x508/0x748 [10919.720711] __arm64_sys_execve+0x40/0x50 [10919.724707] do_el0_svc+0xf4/0x1b8 [10919.728095] el0_sync_handler+0xf8/0x124 [10919.732003] el0_sync+0x140/0x180 Fixes: f9d81a657bb8 ("coresight: perf: Allow tracing on hotplugged CPUs") Reported-by: Jeremy Linton <jeremy.linton(a)arm.com> Cc: Mathieu Poirier <mathieu.poirier(a)linaro.org> Cc: Mike Leach <mike.leach(a)linaro.org> Signed-off-by: Suzuki K Poulose <suzuki.poulose(a)arm.com> --- drivers/hwtracing/coresight/coresight-etm-perf.c | 10 ++++++++++ 1 file changed, 10 insertions(+) diff --git a/drivers/hwtracing/coresight/coresight-etm-perf.c b/drivers/hwtracing/coresight/coresight-etm-perf.c index 1a3169e69bb1..9d61a71da96f 100644 --- a/drivers/hwtracing/coresight/coresight-etm-perf.c +++ b/drivers/hwtracing/coresight/coresight-etm-perf.c @@ -321,6 +321,16 @@ static void etm_event_start(struct perf_event *event, int flags) if (!event_data) goto fail; + /* + * Check if this ETM is allowed to trace, as decided + * at etm_setup_aux(). This could be due to an unreachable + * sink from this ETM. We can't do much in this case if + * the sink was specified or hinted to the driver. For + * now, simply don't record anything on this ETM. + */ + if (!cpumask_test_cpu(cpu, &event_data->mask)) + goto fail_end_stop; + path = etm_event_cpu_path(event_data, cpu); /* We need a sink, no need to continue without one */ sink = coresight_get_sink(path); -- 2.24.1

4 years, 3 months

[PATCH] coresight: cti: disclaim device only when it's claimed

by Tingwei Zhang

Coresight_claim_device() is called in cti_starting_cpu() only when CTI is enabled while coresight_disclaim_device() is called uncontionally in cti_dying_cpu(). This triggered below WARNING. Only call disclaim device when CTI device is enabled to fix it. [ 75.989643] WARNING: CPU: 1 PID: 14 at kernel/drivers/hwtracing/coresight/coresight.c:209 coresight_disclaim_device_unlocked+0x10/0x24 [ 75.989697] CPU: 1 PID: 14 Comm: migration/1 Not tainted 5.9.0-rc1-gff1304be0a05-dirty #21 [ 75.989709] Hardware name: Thundercomm Dragonboard 845c (DT) [ 75.989737] pstate: 80c00085 (Nzcv daIf +PAN +UAO BTYPE=--) [ 75.989758] pc : coresight_disclaim_device_unlocked+0x10/0x24 [ 75.989775] lr : coresight_disclaim_device+0x24/0x38 [ 75.989783] sp : ffff800011cd3c90 . [ 75.990018] Call trace: [ 75.990041] coresight_disclaim_device_unlocked+0x10/0x24 [ 75.990066] cti_dying_cpu+0x34/0x4c [ 75.990101] cpuhp_invoke_callback+0x84/0x1e0 [ 75.990121] take_cpu_down+0x90/0xe0 [ 75.990154] multi_cpu_stop+0x134/0x160 [ 75.990171] cpu_stopper_thread+0xb0/0x13c [ 75.990196] smpboot_thread_fn+0x1c4/0x270 [ 75.990222] kthread+0x128/0x154 [ 75.990251] ret_from_fork+0x10/0x18 Fixes: e9b880581d55 ("coresight: cti: Add CPU Hotplug handling to CTI driver") Signed-off-by: Tingwei Zhang <tingwei(a)codeaurora.org> --- drivers/hwtracing/coresight/coresight-cti.c | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-) diff --git a/drivers/hwtracing/coresight/coresight-cti.c b/drivers/hwtracing/coresight/coresight-cti.c index 3ccc703dc940..d6fea6efec71 100644 --- a/drivers/hwtracing/coresight/coresight-cti.c +++ b/drivers/hwtracing/coresight/coresight-cti.c @@ -742,7 +742,8 @@ static int cti_dying_cpu(unsigned int cpu) spin_lock(&drvdata->spinlock); drvdata->config.hw_powered = false; - coresight_disclaim_device(drvdata->base); + if (drvdata->config.hw_enabled) + coresight_disclaim_device(drvdata->base); spin_unlock(&drvdata->spinlock); return 0; } -- The Qualcomm Innovation Center, Inc. is a member of the Code Aurora Forum, a Linux Foundation Collaborative Project

4 years, 3 months

2024

2023

2022

2021

2020

2019

2018

2017

2016

2015

CoreSight September 2020